fal video models vs Magic Hour MCP

fal video models 77

Graded as a hosted service on the queue API. Instatus page at status.fal.ai with components for Model API, Serverless API, Official Models and the dashboards, and monthly uptime figures (20). The notice history, re-read on 1 November 2026, has two notices in the 90 days from 3 August, a 30-minute major outage of synchronous calls to fal.run on 4 September (the queue wasn't named, so partial) and 18 minutes of degraded performance on 29 September, and none in October (20). Concurrency is published, 2 for a new account rising to 40 self-serve with paid invoices over four weeks, and queued requests are never rejected (15). A 429 carries type concurrent_requests_limit and X-Fal-needs-retry: 1, the docs give backoff of 1, 2, 4 and 8 seconds, the queue retries 503, 504 and connection errors up to 10 times, and backup domains are documented. There is no idempotency key, and one page says concurrency requeues have no maximum while another says 10 (12). The enterprise page mentions SLA guarantees but no SLA document is published (0). The queue API is generally available (10).

Magic Hour MCP 35

Status page at status.magichour.ai, all services online, with maintenance and previous incidents sections; no MCP component read (15). Incident history not read (5). No API rate limits found (0). No retry guidance found; wait_for_* tools poll long jobs (5 of 15). No SLA found (0). Not labelled beta (10). 15 + 5 + 0 + 5 + 0 + 10 = 35.

fal video models 73

Keys take one of seven permission presets (API, SERVERLESS, DEPLOY, COMPUTE, BILLING, READONLY, FULL) or a custom permission list, can be limited to named models and can expire after 1 hour to 1 year, since 9 October 2026. Legacy ADMIN and API keys still work. The MCP relay uses OAuth with browser sign-in, and its scopes aren't documented (30). A READONLY preset grants no write access, and model limits narrow a key further. The MCP docs state that the server has no server-side approval gate before a paid generation (15). Returns generated media and fal's own catalogue text (10). Request payloads are kept 30 days in the dashboard history, and usage and analytics APIs report per-request spend (10). https://fal.ai/.well-known/security.txt returned 404 on 1 November 2026 and no disclosure policy or bug bounty was found. The enterprise page claims SOC 2 certification, the DPA of 31 July 2026 has a data security exhibit, and the trust centre at trust.fal.ai renders only in a browser, so the report and its scope were unread, per the 8 October check (8).

Magic Hour MCP 45

API keys as Bearer tokens or OAuth through the Claude and ChatGPT connectors (20 of 30). Destructive deletes are annotated; no read-only key (8 of 20). Tools return generated media, not third-party text (10). No audit log found (0). No security.txt or certifications found; face and voice tools come with policy limits in the terms (7 of 20). 20 + 8 + 10 + 0 + 7 = 45.

fal video models 81

Graded on the queue API, with the hosted MCP server noted. Results are small JSON with a file URL, and a model's llms.txt is 6 to 8 KB loaded on demand. The MCP server has 11 tools with schema and pricing lookups (20). The catalogue API takes category, limit and a cursor, queue status reports position, webhooks replace polling, and most models take duration and resolution, though names and types differ (18). Errors carry a machine-readable error_type, the same value in X-Fal-Error-Type, and guidance on which to retry (18). No idempotency key. Server-side retries, X-Fal-No-Retry, a cancel endpoint and no charge for 5xx failures reduce the cost of a failed call, but a repeated submit is a second billable job, and the MCP docs don't mention readOnlyHint or destructiveHint (10). Few required parameters, and official Python and JavaScript clients plus Swift, Kotlin and Dart (15).

Magic Hour MCP 55

50 tools totalling about 160 KB of definitions (5). list_saved_items and project getters; limited size controls (10 of 20). Error handling not documented (5 of 20). Every tool annotated, 14 read-only and the three deletes marked destructive (20). Few required inputs and SDKs (Python, Node and others listed in the docs) (15). 5 + 10 + 5 + 20 + 15 = 55.

fal video models 89

An OpenAPI 3.1 document for the queue API and one for the platform APIs, and a per-endpoint OpenAPI 3.1 document from the catalogue API with expand=openapi-3.0, read for fal-ai/kling-video/v3/pro/text-to-video on 1 November 2026 (25). The 8 October reading of per-endpoint documents on fal.ai's ?endpoint_id= path is struck, because fal.ai's robots.txt disallows that query. A docs llms.txt, Markdown twins of docs pages and an llms.txt per model with schema, defaults and price (10). Model pages state what each model does but rarely when not to use it, and the Seedance 2.5 catalogue text says up to 720p while its schema lists 1080p (12). Inputs are typed with enums, ranges and defaults, but duration is a string enum on Kling, a string with a unit on Veo 3.1 and an integer on Wan 3.0, and Kling's rule that one of prompt or multi_prompt is needed sits in prose with a required example of {} (13). Examples in cURL, Python and JavaScript on every model, a model error reference with typed errors such as content_policy_violation and video_duration_too_long, and a request error reference with 13 types (14). Versioned endpoint IDs and a dated public changelog (15).

Magic Hour MCP 82

All 50 tools read from tools/list with JSON Schema inputs and 29 enums, built from Magic Hour's OpenAPI specification (25). llms.txt (10). Long descriptions with a get_workflow_guide tool for choosing between them (17 of 20). Typed inputs with enums and no free-form objects (15). Worked examples; error responses not documented (5 of 15). Dated changelog with entries on 3 to 5 November 2026 (10 of 15, product changelog rather than MCP versions). 25 + 10 + 17 + 15 + 5 + 10 = 82.

fal video models 81

The changelog's newest entry is 29 October 2026 (fallback chains), and the catalogue shows video endpoints added on 16 and 20 October (30). 11 dated changelog entries from 3 August to 29 October 2026 (20). Public changelog, Discord and a support address, with response times not checked (10). Python client at 1.0.3 (repository tag fal_client_v1.0.3, 18 September 2026; PyPI not re-read on 1 November), JavaScript client at 1.10.1 (tag client-v1.10.1, 4 May 2026), and Swift, Kotlin and Dart clients documented (13). fal-js had 19 open issues and pull requests per the 8 October check, not re-read on 1 November, and its CI wasn't read (8).

Magic Hour MCP 75

Changelog entries on 3, 4 and 5 November 2026 (30). More than three dated entries in 90 days (20). Changelog and support; issues on the public MCP repository not read (10 of 25). Registry entry io.github.magichourhq/magic-hour (0.1.1, 4 November) (15). No package health read (0). 30 + 20 + 10 + 15 + 0 = 75. Registry presence read from Anchor Terminal's own daily import of the official MCP registry (/api/v1/registry.json, 6 November 2026), not the registry's API.

fal video models 64

Closed platform with published terms (8 September 2026) and API Services supplemental terms. The clients are MIT and the catalogue marks each model's license_type. The terms warrant nothing about rights in output, and a clause bars scripts that integrate with the services without written authorisation, which sits oddly with an API product (13). Data handling is documented and mostly agrees. Request payloads are kept 30 days with an opt-out header and a delete API, the API Services terms say client content isn't used to train or develop fal's products except on models marked Pending Enterprise Ready, and the privacy policy of 22 July 2026 deletes account data 30 days after closure. The terms still let fal use usage data to develop AI models, and content sent to a third-party model goes to that third party (22). Model pages now carry dated deprecation notices, Veo 3.1 and its sub-endpoints on 17 November 2026, Luma Ray 2 on 25 October with a named replacement, and the key docs say presets can be refined. But 31 of 152 text-to-video endpoints are marked deprecated in the catalogue, most without a date, and there is no written policy or changelog entry (12). The DPA names a subprocessor list at trust.fal.ai/subprocessors with 15 days' notice of changes, but that page renders only in a browser, and the privacy policy says processing is in the United States and other countries (10).

Magic Hour MCP 58

Closed service under terms naming Magic Hour AI, Inc., revised 3 October 2026 (15). The privacy policy says prompts, uploads and outputs aren't sold or used to train models, for free and paid accounts, and MCP requests are handled the same way; retention periods aren't given (20 of 30). No deprecation policy found (5 of 20). Subprocessors not read (10 of 20, model providers named in tool docs). 15 + 20 + 5 + 10 = 50.